Verifiable computation outsourcing has become a critical tool in various applications, particularly in scaling blockchain networks like Ethereum. A prominent example is the use of rollups, a layer-2 scaling solution designed to increase Ethereum’s transaction throughput. In this approach, transactions are offloaded to a secondary layer that mirrors Ethereum’s functionality but operates on a high-performance cluster. To maximize efficiency, hundreds of transactions are batched, compressed, and processed together. The resulting state of this layer-2 ledger is then computed and submitted to the Ethereum mainnet for verification and permanent record-keeping.
A key component of this verification process is the use of Zero-Knowledge Virtual Machines (zkVMs), which enable efficient and privacy-preserving proof generation. Currently, there are approximately a dozen team actively developing zkVM implementations. Given the growing interest and diversity in zkVM designs, our objective is to establish a comprehensive set of criteria for objectively evaluating and comparing their performance.